Shared and distinctive microbial biomarkers enhance precision diagnostics of inflammatory bowel disease and its subtypes

Inflammatory Bowel Disease (IBD) is a chronic inflammatory gut disease with major subtypes of Crohn's disease (CD) and ulcerative colitis (UC). They significantly differ in pathology but are similar in clinical manifestations, bringing considerable challenges for accurate non-invasive diagnosis. Recent research highlights the gut microbiome as a promising physiological indicator for diverse diseases, prompting us to investigate shared or distinctive microbial species insights into IBD and its subtypes. For this purpose, we initially established a standardized computational workflow to identify gut microbial biomarkers and their patterns for IBD diagnosis and subtype discrimination. In the process, 1,210 metagenomic samples were analyzed from 6 cohorts; among them, 81 shared species were identified (18 enriched, 63 depleted) across both CD and UC, whereas CD harbored 73 distinct species (37 enriched, 36 depleted) and UC had 32 distinct species (19 enriched, 13 depleted). Functional profiling illustrated shared and distinctive enriched MetaCyc pathways across the two subtypes, offering a comprehensive view of subtype-specific metabolic pathways. Finally, a two-stage machine learning framework was built for diagnosis and subtyping of IBD, making IBD identification in the first stage and subtype classification in the second stage. The framework was well-trained and achieved considerable accuracy, even leading to a novel discovery that shared microbes dominate IBD diagnosis, while distinct microbes drive subtype discrimination.